Genetical analysis of the F2 triple test cross design combined with conventional early generations was used to elucidate the genetical control of yield and yield components in two crosses of winter wheat. From estimates of the additive, {d}, and additive X additive, {i}, components of means, together with the additive genetical variance, D, predicted frequencies of recombinant inbred lines that would transgress the parental range were calculated for each cross. The accuracy of predictions was evaluated by comparing expected frequencies with observed numbers in populations of F6 lines previously developed by single seed descent.For both crosses and all characters where an adequate genetical model was found to explain the observed variation between the early generations, good agreement between predicted and observed frequencies of transgressive segregants was obtained. Furthermore, for characters exhibiting significant epistasis, allowance for additive X additive {i} epistasis in the prediction equations was sufficient to allow for skewness of the recombinant inbred population.These results demonstrate that cross performance in wheat can be predicted from genetical analysis of early generations, and the value of this approach in breeding new varieties is discussed.